| Age | Commit message (Collapse) | Author |
|---|
|
Updated security policy
|
|
Co-authored-by: Codex <[email protected]>
|
|
Merging changes for the v.6.5.1.202602 release
|
|
Applied the standard MIT license header to all project-owned C, header,
assembly, shell, and Python files that were missing a copyright notice.
Third-party, toolchain startup, and auto-generated files were excluded.
Co-authored-by: Copilot <[email protected]>
|
|
* Updated version number constants
* Updated port version strings
---------
Co-authored-by: Copilot <[email protected]>
|
|
Co-authored-by: Copilot <[email protected]>
|
|
client lifecycle (#265)
PR #262 introduced per-instance UX_HOST_CLASS_HID_CLIENT allocation in
client_search.c. However, keyboard/mouse/remote_control activate handlers
already embed a UX_HOST_CLASS_HID_CLIENT inside their own combined
allocation (e.g. UX_HOST_CLASS_HID_CLIENT_KEYBOARD), override
hid->hid_client with the embedded copy, and then free the entire combined
struct (as 'keyboard_instance', the first field) during deactivation.
This created two bugs:
1. Memory leak: the per-instance copy from client_search was abandoned
when activate handlers replaced hid->hid_client with their embedded
copy.
2. Double-free / UX_MEMORY_CORRUPTED: deactivate.c freed hid->hid_client
after calling the handler, but for keyboard/mouse/remote_control the
deactivate handler had already freed the entire combined allocation
(which contains the embedded hid_client), causing a second free of a
pointer into the middle of a now-freed block.
Fix:
- keyboard/mouse/remote_control activate: free the per-instance copy from
client_search before overriding hid->hid_client with the embedded one.
- keyboard/mouse/remote_control deactivate: null hid->hid_client after
freeing the combined struct, signalling that cleanup is done.
- deactivate.c: re-check hid_client != NULL after calling the handler
before freeing; keyboard/mouse/remote_control will have nulled it,
simple clients will not.
- keyboard/mouse ACTIVATE_WAIT error paths (standalone): null
hid->hid_client after freeing the combined struct so that the generic
cleanup in entry.c skips the already-freed pointer.
- entry.c standalone ACTIVATE_WAIT error: guard the free with a NULL
check to safely handle both cases.
Discovered while investigating test failures introduced by PR #262.
All 430 tests pass after this fix.
Co-authored-by: Copilot <[email protected]>
|
|
* fix(hid): Store client instance per-device instead of per-class
* Address review feedback on per-instance HID client copy
Three issues fixed, discovered during maintainer review:
1. Memory leak in standalone activation error path (entry.c):
_ux_host_class_hid_client_activate_wait() set hid_client to NULL
without freeing the per-instance copy allocated in client_search.
The HID_ENUM_ERROR handler destroys the hid struct without freeing
hid_client, so the copy was leaked on every standalone activation
failure. Fixed by freeing hid_client before clearing it.
2. Variable declared inside if-block (client_search.c):
hid_client_instance was declared inside the if (status == UX_SUCCESS)
block, which is a C99 feature. USBX targets C89/C90 embedded
toolchains. Moved to the top of the function with other locals.
3. Trailing whitespace throughout both changed files:
The PR introduced trailing spaces on most comment-block lines.
Reverted all affected lines to their original whitespace.
Co-authored-by: Copilot <[email protected]>
---------
Co-authored-by: Frédéric Desbiens <[email protected]>
Co-authored-by: Copilot <[email protected]>
|
|
Two issues discovered while reviewing and fixing thread-safety issues in
the new ux_host_class_hid_protocol_get/set functions (PR #244).
1. idle_get.c: wrong operator acquires no lock in standalone mode
Line 104 used '&= ~UX_HOST_CLASS_HID_FLAG_LOCK' (the release/clear
operation) instead of '|= UX_HOST_CLASS_HID_FLAG_LOCK' (acquire/set).
The preceding check correctly returns UX_BUSY when the flag is set,
but the follow-on line then immediately clears it instead of setting it.
The net effect is that the HID instance is never actually locked in
UX_HOST_STANDALONE mode, making the mutual-exclusion check a no-op.
2. idle_set.c: standalone path used a blocking spin-loop
The standalone branch called _ux_host_class_hid_idle_set_run() in a
do/while loop, blocking the caller until the transfer completed. This
is inconsistent with every other inline HID control-transfer function
(idle_get, report_get, report_set, protocol_get, protocol_set) which
all use the direct UX_DISABLE/UX_RESTORE atomic flag pattern and
return UX_BUSY if the instance or device endpoint is already locked.
Replaced with the same inline standalone locking pattern used by the
other functions: atomic HID FLAG_LOCK acquire, atomic DEVICE_FLAG_LOCK
acquire with AUTO_DEVICE_UNLOCK + UX_TRANSFER_STATE_RESET, and an
AUTO_WAIT check at completion consistent with idle_get behavior.
Co-authored-by: Copilot <[email protected]>
|
|
get/set API (#244)
This pull request adds host-side HID API functions to get and set the HID protocol (boot vs. report mode), and introduces an optional device-sidecallback invoked when the protocol changes.
New features
- Added host-side APIs ux_host_class_hid_protocol_set() and ux_host_class_hid_protocol_get() with error-checking wrappers (_uxe_ variants).
- Added optional device-side callback ux_device_class_hid_set_protocol_callback, invoked when the host changes the active protocol.
- Added a comprehensive test to verify the new protocol callback functionality.
Bug fixes (host-side protocol get/set)
1. DMA buffer safety (protocol_get): The received byte is now written into a cache-safe allocated buffer instead of the caller's USHORT* directly. Stack memory is not DMA-safe on cache-incoherent embedded targets, and writing 1 byte into a USHORT* yields incorrect results on big-endianplatforms. Follows the same pattern as ux_host_class_hid_idle_get.
2. Missing device protection semaphore (both functions): In RTOS mode, ux_device_protection_semaphore is now acquired before submitting the transfer, in addition to the HID instance semaphore. All other HID control transfer functions (idle_get, idle_set) take both semaphores; omitting it allowed concurrent callers to corrupt the shared transfer_request fields.
3. Missing standalone locking (both functions): Added proper UX_HOST_STANDALONE support using UX_DISABLE/UX_RESTORE to atomically check and setUX_HOST_CLASS_HID_FLAG_LOCK and UX_DEVICE_FLAG_LOCK, and set UX_TRANSFER_FLAG_AUTO_DEVICE_UNLOCK + UX_TRANSFER_STATE_RESET on the transfer request before initiating. Follows the idle_get inline standalone pattern.
Co-authored-by: Frédéric Desbiens <[email protected]>
Co-authored-by: Copilot <[email protected]>
|
|
* No longer using request_value to get request type.
* Removal of shift.
* Fix GET/SET_DESCRIPTOR routing for class-defined descriptor types
When a host sends GET_DESCRIPTOR or SET_DESCRIPTOR with bmRequestType
set to STANDARD but requests a class-defined descriptor type (e.g. HID
Report 0x22 or Physical 0x23), the request must be routed to the class
layer rather than handled as a standard USB request.
Per USB HID 1.11 section 7.1.1, HID stacks legitimately issue
GET_DESCRIPTOR with bmRequestType=STANDARD | INTERFACE | IN for class
descriptors. The previous fix (checking request_type != STANDARD) broke
this path: (0x81 & 0x60) == 0x00 was seen as standard and the request
would be stalled.
The new condition explicitly checks:
1. request is GET_DESCRIPTOR or SET_DESCRIPTOR
2. bmRequestType type field is STANDARD
3. bDescriptorType (high byte of wValue) is in the USB-IF class-reserved
range 0x21..0x2F
Requests with bmRequestType already set to CLASS or VENDOR, and standard
descriptors (bDescriptorType < 0x21, e.g. BOS 0x0F) or vendor descriptors
(>= 0x40), are left unchanged and follow their normal dispatch path.
Also fix (UINT) to (ULONG) cast, matching the declared type of request_type.
Suggested-by: ABOUSTM <https://github.com/ABOUSTM>
---------
Co-authored-by: Frédéric Desbiens <[email protected]>
Co-authored-by: Copilot <[email protected]>
|
|
This commit cleans up header declarations by removing API and test prototypes that have no corresponding implementation in the repository.
Impact:
- No runtime behavior change.
- Compile/link surface is cleaner and more accurate for users and tests.
|
|
Added default value protection for line coding parameters in serial Host classes
|
|
Added missing uninitialize macro in simulator host header (HOST SIM)
|
|
Add preprocessor guards to allow customization of default line coding parameters.
Summary of Changes:
- Wrapped default line coding constants (RATE, STOP_BIT, PARITY, DATA_BIT)
with #ifndef guards in CDC ACM, GSER, and Prolific host class headers
- This allows users to define custom default values before including these headers in (ux_user.h)
- Updated Prolific activation code to use DEFAULT constants instead of hardcoded values
for consistency and better maintainability
Files Modified:
- common/usbx_host_classes/inc/ux_host_class_cdc_acm.h
- common/usbx_host_classes/inc/ux_host_class_gser.h
- common/usbx_host_classes/inc/ux_host_class_prolific.h
- common/usbx_host_classes/src/ux_host_class_prolific_activate.c
|
|
Merge changes ahead of the v6.5.0.202601 release
|
|
Added break request support to Device CDC ACM
|
|
|
|
|
|
Pre v6.5.0.202601 fixes
|
|
|
|
ux_host_class_hid_report_item_analyse
|
|
Updated copyright headers and version number constants
|
|
|
|
|
|
|
|
Improved code to leverage the ux internal macros instead of using direct tx struct fields
|
|
Added START STOP (load/eject) request support to Device MSC
|
|
Fixed Linux enumeration for device RNDIS
|
|
|
|
Added uninitialize API to the DCD Simulator Controller Driver
|
|
BerninaInternationalAG/create_mutex_before_memory_allocation
Modified logic to create mutex before calling _ux_utility_memory_allocate
|
|
When compiling with UX_ENABLE_DEBUG_LOG the call to
_ux_utility_memory_allocate requires the system mutex.
|
|
Added a check for NULL packet before using the pointer
|
|
Removed duplicated defines
|
|
- Implements handling of UX_SLAVE_CLASS_CDC_ACM_SEND_BREAK in the CDC ACM control request path.
- Ensures break state is cleared on class deactivation to avoid carrying state across disconnect/reset cycles.
- Minor comment/whitespace cleanups in touched headers/sources.
- Testing: Build-only / compilation sanity/send break request. (no new automated tests added).
|
|
|
|
- Adds _ux_dcd_sim_slave_uninitialize() to free the simulated slave DCD controller instance, clear DCD bindings, and return the DCD to UX_UNUSED (idempotent if already unused).
- Registers the new uninitialize source in CMakeLists.txt.
- Updates simulator headers to export ux_dcd_sim_slave_uninitialize (and ux_hcd_sim_host_uninitialize) and does minor whitespace/style cleanup in the simulator header blocks.
|
|
- Add mass-storage “removable media” support by tracking per-LUN prevent/allow medium removal and loaded/ejected status.
- Introduce an optional per-LUN callback to handle SCSI START STOP UNIT (including load/eject and power-condition fields).
- Extend storage constants (media types, sense keys/codes, power conditions, prevent flags) to support the new behavior and improve readability (explicit hex values).
- Initialize new per-LUN state on activation (default: medium removal allowed, medium loaded) and wire the new callback through storage initialization.
- Includes minor formatting/comment cleanups and version banner updates to 6.4.6.
SCSI Block Commands – 4 (SBC-4)
|
|
Added ports for RXv1/GNU, RXv2/GNU and RXv3/GNU
|
|
https://github.com/ayedm1/eclipse_usbx into refer_to_ux_internal_defines
|
|
|
|
|
|
If the queue is empty we need the check here.
|
|
Merging changes ahead of hotfix release 202504a
|
|
|
|
Fixed CI build issues
|
|
- Issue due to "fix (host/storage): prevent stack overflow from infinite partition recursion" commit
Fix
- Add missing UX_HOST_CLASS_STORAGE_MAX_PARTITIONS_COUNT in some ux_port.h
|
|
|
|
Merge changes ahead of the 202504 release
|
